Experiment Preferences jobs-apply

Deploying 4 parallel Claude Code agents in isolated git worktrees can divide adapter development work and merge results into a clean baseline

ai-agentsorchestrationcareer
Hypothesis

Deploying 4 parallel Claude Code agents in isolated git worktrees can divide adapter development work and merge results into a clean baseline

Result: confirmed
Key Findings

679 sessions (522 main + 157 worktree) in 2 days. Agents successfully parallelized adapter work. Merged into baseline commit e4f43c5. Pattern later formalized in Dakka orchestrator.

Changelog

DateSummary
2026-04-07Created during temporal gap audit
2026-03-09Original experiment

Hypothesis

Deploying 4 parallel Claude Code agents in isolated git worktrees : each assigned to a specific ATS channel : can divide adapter development work and produce mergeable results faster than sequential development.

Method

4 Opus sub-agents assigned to: Direct Discovery, Direct Application, LinkedIn Worker, Greenhouse Worker. Each operated in an isolated worktree branched from the same commit. The orchestrator session monitored progress and merged changes. 679 total sessions across 9 worktree directories over 2 days.

Results

Confirmed. The parallel work produced adapter improvements that merged cleanly into baseline commit e4f43c5. The pattern demonstrated that Claude Code sessions could be orchestrated as parallel workers with minimal merge conflicts when given separate code domains.

What Carried Forward

This was the foundational experiment that led to the Dakka parallel orchestrator project. The insight : “each agent gets its own worktree and its own code domain” : became Dakka’s core architecture.